Since not many linguists tend to frequent it, I thought I’d share some interesting stuff from the latest Cognitive Science Society conference.
Although the society is not well-known for its friendliness to linguists (esp. with its high registration fees), at this year’s meeting Paul Smolensky was awarded the field’s highest distinction (the Rumelhart Prize). There were also several posters on phonology (two from Hopkins):
- Sara Finley (JHU) presented an interesting poster detailing on OT analysis of morphologically-conditioned vowel harmony (abstract link; a more detailed handout is here). Unfortunately, the data I found most interesting are not mentioned in either document. Apparently, the Korean morphologically-conditioned harmony process violates the structure-preserving generalization that vowels are neutral only when harmonizing would create a vowel outside of the inventory. She also has some data from a Spanish dialect (suggesting the complementary case, where morphologically-conditioned harmony creates vowels that are otherwise absent from the inventory.
- Adam Wayment (JHU) presented a poster on getting harmony networks (a type of connectionist network implementing a ‘soft’ version of OT) to learn basic syllable structure constraints (abstract). I can’t say I really had enough time to grasp the details of this.
- Brent Vander Wyk and James McClelland (CMU) presented a poster detailing work that they attempted to model the frequency distribution of English rhymes using a grammar with numerically weighted constraints. They postulated a set of constraints, loosely inspired by articulatory considerations, and used linear regression to assign weights to the constraints. The fit they got was ok, but not shockingly impressive.
In sum, the conference’s connections to linguistics, and phonology in particular, are still rather weak. It’s my hope that Smolensky’s prize award will help improve this situation.